Effects of the liver volume and donor steatosis on errors in the estimated standard liver volume

Authors


Abstract

An accurate assessment of donor and recipient liver volumes is essential in living donor liver transplantation. Many liver donors are affected by mild to moderate steatosis, and steatotic livers are known to have larger volumes. This study analyzes errors in liver volume estimation by commonly used formulas and the effects of donor steatosis on these errors. Three hundred twenty-five Asian donors who underwent right lobe donor hepatectomy were the subjects of this study. The percentage differences between the liver volumes from computed tomography (CT) and the liver volumes estimated with each formula (ie, the error percentages) were calculated. Five popular formulas were tested. The degrees of steatosis were categorized as follows: no steatosis [n = 178 (54.8%)], ≤10% steatosis [n = 128 (39.4%)], and >10% to 20% steatosis [n = 19 (5.8%)]. The median errors ranged from 0.6% (7 mL) to 24.6% (360 mL). The lowest was seen with the locally derived formula. All the formulas showed a significant association between the error percentage and the CT liver volume (P < 0.001). Overestimation was seen with smaller liver volumes, whereas underestimation was seen with larger volumes. The locally derived formula was most accurate when the liver volume was 1001 to 1250 mL. A multivariate analysis showed that the estimation error was dependent on the liver volume (P = 0.001) and the anthropometric measurement that was used in the calculation (P < 0.001) rather than steatosis (P ≥ 0.07). In conclusion, all the formulas have a similar pattern of error that is possibly related to the anthropometric measurement. Clinicians should be aware of this pattern of error and the liver volume with which their formula is most accurate. Liver Transpl, 2011. © 2011 AASLD.

Before the era of liver transplantation (LT), the standard liver volume was mostly estimated for the calculation of drug doses. With the increasing popularity of LT, an accurate assessment of the liver volume was deemed necessary.1 This assessment is more important in living donor LT because of the need for accurate calculations of the safe residual volume for the donor and the optimum volume for the recipient to prevent small-for-size syndrome. The ideal formula for estimating the standard liver volume has been a matter of debate. Many centers have developed their own formulas to represent the populations that they serve.2-6 Most of the formulas use the body weight (BW), body height (BH), or body surface area (BSA) as the variable parameter.

Hepatic steatosis is seen in half of the donors for living donor LT.7, 8 The impact of hepatic steatosis on living donor LT has been widely discussed.9, 10 Steatotic grafts have been clearly documented to have larger volumes.11, 12 However, none of the previously published studies have assessed the accuracy of the formulas with extreme liver volumes or in the presence of hepatic steatosis. In this study, we analyzed the errors in the estimation of the standard liver volume and the effects of steatosis on these errors in a cohort of Asian living liver donors who underwent right lobe donor hepatectomy.

Abbreviations: BH, body height; BSA, body surface area; BW, body weight; CT, computed tomography; ESLV, estimated standard liver volume; G0, group 0; G1, group 1; G2, group 2; LT, liver transplantation.

PATIENTS AND METHODS

Donors who underwent right liver donor hepatectomy (with the middle hepatic vein included) from 1999 to 2009 were the subjects of this study. This study was done on previously collected data in a computer database. No ethical clearance or institutional approval was required. Non-Chinese donors were excluded. In all, 325 donors were included. All of them underwent a preoperative liver volume assessment. The estimation of the liver volume with computed tomography (CT) was performed with the Heymsfield method. All subjects underwent single-slice spiral CT (HiSpeed Advantage System, General Electric Healthcare, Milwaukee, WI) and multislice CT studies (LightSpeed QX/i 4-MDCT or 16-MDCT scanner, General Electric Healthcare). Cuts were made continuously with intervals of 5 to 7.5 mm. The demarcation of the right and left portions of the liver was made via tracing along the middle hepatic vein, which corresponded to Cantlie's line.

The error of the estimated standard liver volume (ESLV) was defined as the difference between the volume estimated with a certain formula and the volume calculated with CT volumetry (the estimated volume minus the CT volume). For the analysis, the error was expressed as a percentage of the estimated volume. Five formulas were tested, and they were selected because they use different anthropometric measurements and different mathematical functions for calculating ESLV:

equation image

For the BSA estimation, the formulas of Urata et al. and Lee et al. use the equation of Du Bois and Du Bois13:

equation image

The formulas of Yoshizumi et al. and Vauthey et al. use Mosteller's equation14:

equation image

For the conversion of the hepatic mass in grams to the hepatic volume in milliliters, a conversion factor of 1.19 was used.2

The second part of the analysis was performed to assess the effect of hepatic steatosis on the estimation error. Twenty-one of the 325 donors (6.5%) underwent preoperative biopsy. Liver biopsy was performed for donors with a body mass index of 27 to 30 kg/m2 who did not have obvious CT evidence of fatty changes in the liver and for donors who had unexplained hyperbilirubinemia or an unexplained elevation of the aspartate aminotransferase or alanine aminotransferase level. All donors underwent core biopsy of the graft intraoperatively before wound closure. The core of the tissue was examined with hematoxylin and eosin staining. The donors were divided into 3 groups according to their degree of macrovesicular steatosis. The degrees of steatosis were categorized as follows: no steatosis [group 0 (G0), n = 178 (54.7%)], ≤10% steatosis [group 1 (G1), n = 128 (39.3%)], or >10% to 20% steatosis [group 2 (G2), n = 19 (6%)]. In G2, 4 donors had >20% steatosis (3 with 25% steatosis and 1 with 30% steatosis), which was detected by routine postimplantation biopsy. The baseline features of the 3 groups are shown in Table 1.

Table 1. Baseline Parameters of the 3 Steatosis Groups
ParameterG0 (n = 178)G1 (n = 128)G2 (n = 19)P Value*
  • NOTE: G0 patients had no steatosis, G1 patients had ≤10% steatosis, and G2 patients had >10% to 20% steatosis.

  • *

    Kruskal-Wallis test.

Age (years)34 (18-58)34.5 (18-58)41 (22-47)0.09
BW (kg)52 (41-74)60 (39-95)63.5 (43.5-84)<0.001
Body mass index (kg/m2)20.26 (16.5-29)22.7 (16-32)24.4 (18-28.6)0.001
Remaining left liver (%)36 (23.6-49.5)34.6 (26-47.7)33.5 (26-42)0.001
Graft weight (g)565 (320-890)637.5 (390-1020)670 (550-1140)<0.001

Statistical Analysis

Linear regression analysis was performed to depict the relationship between the error and the CT liver volume (Fig. 1). The donor body mass index, BSA, BW, degree of steatosis, and CT liver volume were used as variables in a logistic regression analysis of the factors affecting ESLV. The BSA, degree of steatosis, and CT liver volume were included as significant factors in the multivariate model. For the comparison of medians across the groups, the Kruskal-Wallis test was used. SPSS 15.0 (SPSS, Chicago, IL) was used for the statistical analysis.

Figure 1.

Graphs showing the regression lines for the error percentage versus the CT liver volume (mL). All 5 formulas showed a significant correlation between the error percentage and the CT liver volume (P < 0.001). The r2 values are shown in the graphs.

RESULTS

The mean CT liver volume was 1098 mL (724-1746 mL). Table 2 summarizes the errors of the 5 tested formulas. The median error was smallest with Chan et al.'s formula2 (7 mL or 0.6%), and this was followed by the formulas of Urata et al.3 (21 mL or 1.9%), Lee et al.6 (89 mL or 7.4%), Vauthey et al.5 (127 mL or 10.4%), and Yoshizumi et al.4 (360 mL or 24.6%). The formulas of Vauthey et al. and Yoshizumi et al. mostly overestimated the liver volume. With Vauthey et al.'s formula, 66% of the cases had an error greater than 5%. With Yoshizumi et al.'s formula, 96% of the cases had an error greater than 5%. With Urata et al.'s formula (just like Chan et al.'s formula), the number of cases with an error exceeding the 5% upper limit and the number of cases with an error exceeding the −5% lower limit were similar.

Table 2. Error Percentages of the Different Formulas
FormulaMedian Estimated Volume (mL)Median Error (mL)Error (%)Cases Below the −5% Error Level [n (%)]Cases Within the ±5% Error Range [n (%)]Cases Above the 5% Error Level [n (%)]
Chan et al.21110 (801-1908)7 (−501 to 485)0.6 (−63 to 25.4)100 (31)116 (36)109 (34)
Urata et al.31127 (910-1588)21 (−540 to 324)1.9 (−59 to 20.4)88 (27)109 (34)128 (39)
Yoshizumi et al.41462 (1155-2121)360 (−162 to 760)24.6 (−14 to 36)2 (1)10 (3)313 (96)
Vauthey et al.51223 (800-2132)127 (−363 to 726)10.4 (−45.4 to 34)40 (12)69 (21)216 (66)
Lee et al.61195 (982-1646)89 (−473 to 396)7.4 (−48 to 24)42 (13)92 (28)191 (59)

Figure 1 shows a linear regression analysis of the error percentages of the 5 formulas versus the CT liver volume. A significant association (P < 0.001) was noted between the 2 parameters with all 5 formulas. The regression lines show a similar pattern of error with increasing liver volume. The pattern of error shows a stronger relationship with the formulas of Lee et al.6 (r2 = 0.7), Urata et al.3 (r2 = 0.64), and Yoshizumi et al.4 (r2 = 0.6) versus the formulas of Chan et al.2 (r2 = 0.3) and Vauthey et al.5 (r2 = 0.37).

Figure 2 shows that the positive error diminished and the negative error increased with increasing liver volume. With the formulas of Chan et al.2 and Urata et al.,3 the least error was noted when the liver volume was 1001 to 1250 mL. With the formulas of Vauthey et al.5 and Lee et al.,6 the least error was noted when the volume was 1251 to 1500 mL. With Yoshizumi et al.'s formula,4 the error was minimal when the volume was greater than 1500 mL.

Figure 2.

Median error percentages of each formula with different liver volumes: (1) <750, (2) 750 to 1000, (3) 1001 to 1250, (4) 1251 to 1500, and (5) 1501 to 1750 mL.

In phase 2 of the analysis, the effect of hepatic steatosis on the estimated liver volume was assessed. Table 3 shows the median error of each formula for each of the 3 steatosis groups (G0, G1, and G2). With the formulas of Chan et al.2 [G0, 0.56%; G1, 0.65%; and G2, −0.13% (P = 0.99)] and Vauthey et al.5 [G0, 10%; G1, 10%; and G2, 11% (P = 0.97)], no significant difference in error was noted between the 3 groups of steatosis. However, with the formulas of Lee et al.6 [G0, 9%; G1, 6%; and G2, 3% (P = 0.002)], Yoshizumi et al.4 [G0, 25%; G1, 23%; and G2, 21% (P = 0.03)], and Urata et al.3 [G0, 3.6%; G1, 0.5%; and G2, −2% (P = 0.006)], significant differences in error were noted with an increasing degree of steatosis (Fig. 3). In a multivariate analysis of the factors affecting the estimation error, steatosis was not identified as a significant factor (Chan et al., P = 0.07; Urata et al., P = 0.36; Yoshizumi et al., P = 0.50; Vauthey et al., P = 0.14; and Lee et al., P = 0.43). Instead, the estimation error was significantly affected (P < 0.001) by the actual liver volume and the anthropometric measurement that was used (BW in Chan et al.'s formula and BSA in the others; Table 4).

Table 3. Median ESLV Errors of the Different Formulas According to the Degree of Steatosis
FormulaMedian ESLV Errors (%)P Value*
G0G1G2
  • NOTE: G0 patients had no steatosis, G1 patients had ≤10% steatosis, and G2 patients had >10% to 20% steatosis.

  • *

    Kruskal-Wallis test.

Chan et al.20.56 (−33 to 27)0.65 (−35 to 26)−0.13 (−41 to 24)0.99
Urata et al.33.6 (−30 to 33)0.5 (−41 to 30)−2 (−45 to 18)0.006
Yoshizumi et al.425 (−1 to 48)23 (−6 to 45)21 (−10 to 37)0.03
Vauthey et al.510 (−32 to 35)10 (−30 to 37)11 (−29 to 32)0.97
Lee et al.69 (−23 to 37)6 (−34 to 34)3 (−37 to 23)0.002
Figure 3.

Changes in the proportion of error versus different degrees of fatty change in the formulas of Lee et al.,6 Yoshizumi et al.,4 and Urata et al.3: (0) no fatty change, (1) ≤10% change, and (2) >10% to 20% change.

Table 4. Multivariate Analysis of Factors Associated With Estimation Errors
FormulaP Values
CT Liver Volume (mL)BSA (m2) or BW (kg)Steatosis (%)
Chan et al.2<0.001BW, <0.0010.07
Urata et al.3<0.001BSA, <0.0010.36
Yoshizumi et al.4<0.001BSA, <0.0010.50
Vauthey et al.50.001BSA, <0.0010.14
Lee et al.6<0.001BSA, <0.0010.43

DISCUSSION

All the formulas showed a common pattern of error with our Asian cohort of donors. When the actual liver volume was smaller, there seemed to be an overestimation, whereas with larger liver volumes, there was an underestimation. This underestimation of the liver volume was not related to steatosis. Rather, it was related to the actual liver volume and the anthropometric measurement used to calculate the volume. The data derived from donors is useful in the assessment of recipients.

For our analysis, we used the CT liver volume as the donor's accurate liver volume. Previous data have shown that CT volumetry can be used to predict the liver volume with an error of less than 5%.15, 16 The alternative, the volume derived from a donor hepatectomy specimen, seems inappropriate. Data published by our center have indicated that there is a significant discrepancy between specimen and CT assessments of the right lobe volume.17

The median error of each of the 5 formulas was highly variable. The lowest median error was noted with Chan et al.'s formula.2 This could be due to the fact that this formula was derived from the same set of data from the same cohort of donors. Urata et al.'s formula,3 which was originally derived from a cohort of Japanese individuals, showed a lower median error with these donors. This widely accepted formula has been shown to underestimate the liver volume in Caucasians.18 Conversely, the formulas of Vauthey et al.5 and Yoshizumi et al.,4 which were derived from Caucasians, showed the largest median errors with our donors. Similarly, many centers have pointed out the miscalculation of the liver volume with common formulas and have worked out their own.6, 19-21 This highlights the fact that there is no universal formula for estimating the standard liver volume; rather, results are more accurate if the formula is derived from the population that it serves.

In this study, regardless of the formula used, a common pattern of error was noticed. 45.3% of our donors had some degree of steatosis. On the basis of our previous understanding that steatotic livers have larger volumes,11, 12 it may be argued that steatosis is possibly the cause of the observed pattern of error. However, in the multivariate analysis, donor steatosis was not an individual predictor of the error. During selection, we avoided donors who were expected to have more than 20% steatosis. As a result, only 4 donors were beyond this selection criterion, and it is unlikely that they had a significant influence on the multivariate analysis. In our strictly selected donors, mild steatosis did not seem to have a significant impact on the error that we have described here. This is probably because most of our donors had less than 20% steatosis, and this was inadequate for a significant effect on the error.

Analyzing the volumes of 652 cadaveric livers, Yu et al.21 described a nonlinear relationship between a BSA less than 1.2 m2 and the liver volume. Thus, they proposed a nonlinear model for estimating the liver volume. It seems that the liver volume fails to show a linear relationship with anthropometric measurements for extreme values. Thus, the error that we have described is likely a result of this rather than errors in the formulas themselves. For each formula, the possible error can be predicted on the basis of the liver volume, as shown in Fig. 2.

In summary, all 5 formulas have a similar pattern of error that is possibly related to the anthropometric measurement. Clinicians should be aware of this pattern of error and the liver volume with which their formula is most accurate.

Ancillary